CMOS (Complementary Metal Oxide Semiconductor) technology forms the basis of modem integrated circuits and signal processing electronics. CMOS technology has the advantage it consumes very little power, and therefore a large number of transistors can be packed onto a single chip.
Photodetectors are of increasing importance in the field of telecommunications with the advent of optic fiber communications. It would be very useful to fabricate photodetectors for telecommunications applications on a silicon wafer, for example, using SiGe epitaxial layers. If this were possible, the CMOS signal processing electronics could be integrated with the photodetectors on a single chip. Such photodetectors and any associated electronics could also be combined with Si compatible waveguides (e.g. silicon-on-insulator, Si/SiGe/Si, or silica-on-silicon), to create fully integrated optoelectronic modules.
Silicon-germanium (SiGe) alloys have been used for a number of years in the manufacturing of heterojunction bi-polar transistors with considerable success because the introduction of germanium provides more than a doubling of device speed, while preserving 100% compatibility with existing processing technology. SiGe alloys are also attractive for the monolithic integration of Si photonics with mainstream very large scale integrated (VLSI) technology because the addition of Ge extends the wavelength range of silicon, and SiGe/Si multiquantum wells (MQW's) can be epitaxially grown coherent with the Si substrate presenting an additional degree of freedom in band-gap engineering.
The main limitation of this material system for photonic devices is the strain which restricts the thickness and germanium concentration that can be incorporated into a MQW structure. At the large Ge concentrations (&gt;0.5) needed to extend the absorption wavelength to 1.55 .mu.m, the thickness of one strained layer which can be grown without dislocations is limited to 10-20 nm, depending on growth rate and temperature, but the critical thickness becomes even smaller when more than one strained layer is grown in a MQW structure. Dislocations are generally fatal to electronic device performance and cause a large dark current in photodetectors. It has also been reported that surface rippling (Stranski-Krastanov growth) limits planar strained growth to an even smaller critical thickness at Ge concentrations higher than 0.25. The small SiGe well thickness in turn produces a large quantum confinement shift which pushes the band gap to larger energy and thus the absorption threshold to lower wavelengths. Possible solutions are to grow the MQW layers on relaxed buffer layers, allowing for relatively defect free, thick SiGe layers to be grown, or to incorporate carbon into the layers thus compensating for the lattice mismatch introduced by Ge. However, strain is known to decrease the band-gap energy, and relaxed layers would need to contain even more Ge in order to have a band gap similar to strained material.
Over the last few years, Stranski-Krastanov three-dimensional (3D) growth of SiGe has been used to produce dislocation-free strained material, but applications in optoelectronics are yet to be seen. Only limited data have been published so far showing phonon resolved photoluminescence at low energy (&lt;900 meV) from 3D SiGe growth and the band-gap energy obtained was larger than the predicted value for planar growth, making optoelectronic applications even more difficult. In particular, it has been difficult or impossible to fabricate photodetectors or light emitting devices on Si substrates that are useful in the .lambda.=1550 nm telecommunications wavelength band. The reason for this is that only Si.sub.1-x,Ge.sub.x, alloys with high Ge (germanium) concentrations (x&gt;0.5) have energy gaps small enough to be sensitive to light at .lambda.=1550 nm (0.8 eV). Unfortunately, the large lattice mismatch between Si and Ge means that only very thin Si.sub.1=x Ge.sub.x, layers can be grown epitaxially on Si at high Ge concentrations. Exceeding this critical thickness limit results in the formation of dislocations, which degrade photodetector performance. The quantum confinement effect in such thin layers also moves the optical absorption edge to shorter wavelengths, counteracting the effect of increasing the Ge concentration.
SiGe photodetectors for 1550 nm have been fabricated using relaxed buffer layers under the photodetector structure. These buffer layers are used to localize dislocations to within the buffer layer itself and prevent them from threading to the surface through the active photodetector. The drawback of this approach is that the active SiGe layer is also relaxed and has a higher band gap than would be achieved using fully strained growth. As a result, either thicker layers and/or a higher Ge concentration are required to achieve a usable photodetector response. The buffer layers are also quite thick, and their presence can limit the possibilities for subsequent integration with electronics and optical waveguides.
The only other possibility for integrating photodetectors on Si is to use hybrid integration techniques to place III-V semiconductor detectors on a Si wafer. Although this is technically possible, hybrid integration is a rather expensive manufacturing technology, which is not yet really suited to mass production.
An object of the invention is to provide a method of fabricating a photonic device with enhanced photoresponse at longer wavelengths, in particular 1550 nm, on a silicon substrate.